An application of the generalized weight share method in the French Structural Business Statistics survey
Jul 25, 15:45
The French Structural Business Statistics (SBS) production system, Esane , has two main uses :
- the answer to the SBS European regulation ;
- the estimations of businesses contribution to GDP for the national accounts.
It is based on a mix of exhaustive administrative fiscal data and data obtained on a random sample of the population. Esane has been developed to produce accurate estimates on levels and evolutions. Therefore, SBS surveys design has been a rotating sample renewed by half.
Esane is currently changing to produce estimates based on profiled units or enterprises  and no longer on legal units. Starting in 2016, the sampling design of SBS surveys selects profiled units, but information is still collected on the cluster sample of legal units belonging to the sampled enterprises.
The implementation of a rotating scheme for the new sampling design is still work in progress due to changes in sample allocations to adapt to the new dissemination units. The rotating scheme as well as the whole production process has also to adapt to the changes in profiled units definition, that is changes in the list of legal units they are made of.
The paper will focus on the latter, that is the management of modifications of the linkages between legal units and enterprises. The analyses will focus on the impact of enterprises changes for the current sample between sample selection and data collection and analysis, as a first step, but the methodology will probably be extendable to the management of linkage modifications for a longitudinal survey.
The aim of the study is to compare two variants of the Generalized weight share method  :
- The "classical" Generalized weight share method (each legal unit "equally" contributes to the weight of the enterprise, whatever its economic characteristics);
- The Generalized weight share method with weighted links (the contribution to the weight of the enterprise depends on the economic characteristics (for example the turnover) of the legal unit).
The comparison will be based on simulations of sampling, thanks to administrative data where variables with strong correlation with the SBS variables are available to each legal unit.